Question

I’ve got 2 data frames – learn data with L rows and test data with T rows.

I want to compute a L*T matrix with distances (euclidean, manhattan, cosine…) between according elements.

Here is my take:

distance2 <- function (x1, x2) {
            temp <- x1 - x2
            sum(temp * temp)
}

m <- matrix(0,nrow(learnData),nrow(testData))
for(td in 1:nrow(testData)) {
    for(ld in 1:nrow(learnData)) {
        m[ld,td] <- distance2(testData[td,],learnData[ld,])                 
    }
}

I think this can be done in a more compact, “R” way. Any ideas?
Thanks.

Answer 1

Two options spring to mind:

1. Use the proxy package which has many of these dissimilarity coefficients already coded and can compute this for two data frames separately
2. The analogue package, which has function distance() which can compute the Euclidean and Manhattan measures for you on two data frames (but not the cosine distance).